In the last episode (Aug 26), Daniel Eriksson said: > Robert Watson wrote: > > Could you be more specific about "one way" or "another"? > > Not at this point. The machine is on the other side of town and needs > to be online for the next couple of days. > > The corrupted files are large (10-400MB) binary files, and the > content of the files makes it hard to pinpoint exactly what is > corrupted. > > I will see what I can do about providing more information, but I > cannot make any promises right now. I used the following two programs to help diagnose some file corruption problems in Linux's XFS and NFS code. Genoffsets creates a 1.5gb file called "offsets" with the 4-byte integers 0, 4, 8, 16, etc. in network-byte order (so each number "n" starts at file offset "n"). You than copy that file (or a part of it) to a remote system and run cmpoffsets on it to verify that the file copied ok. Cmpoffsets just reads the file and checks that the contents are correct. For runs of data that are incorrect, it prints the start and end offsets (and whether they're on power-of-2 boundaries). If the corrupted data is zeros, it prints that. Otherwise it assumes the data was copied from another location and prints the start and end offsets of the copied data (and whether they're on power-of-2 boundaries). Run the copy and verify in a loop that exits if cmpoffsets ever returns a nonzero result. -- Dan Nelson dnelson_at_allantgroup.com
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:08 UTC